I've been trying to check that my current application working on
postgres 7.4 will work with 8.0.
I've not checked the application yet but I'm getting a few error messages,
I don't understand and are worring me slightly. I need to get 8.0 to work
because I'm keen to start using the new pitr feature in 8.0 as soon as it
goes stable.
The error messages run as follows
ERROR: tuple concurrently updated
STATEMENT: ANALYZE "public"."streetuniq"
and
ERROR: could not access status of trasaction 393311
DETAIL: could not read file "/database/db/pg_clog/0000" at offset 98304:
Success
STATEMENT: ANALYZE "public"."backedup"
and
ERROR: relation "people" TID 362/32: OID is invalid
STATEMENT: vacuum full verbose analyze;
I'm running pg_autovacuum which caused the first two errors and
ran a vacuum full verbose analyze straght after loading the data. hence
the third message.
I thought pg_autovaccum was going to be built into 8.0 or was that
only a rumor.
I've also read though the pitr documentation and it seams
slightly complicated. I currently backup every 12 hours (3am and 3pm) with
a dump,
So I'm thinking to get a complete pitr backup with full backup
every 12 hours, I need to take a copy of the data directory and store all
the logs from the start of the backup to the end of the next backup to
have a complete log of that period of time.
I'm planning to store my backup on CD with a new set of CDs for
each backup period.
Whats the basic principle to do this and how can I save on the
number of discs I need, My database is 2.8Gb on disk currently after a
fresh load from dump, the dump is 766Mb uncompressed.
Peter Childs
